Ž . Atmospheric Research 51 1999 245–265 The behavior of total lightning activity in severe Florida thunderstorms Earle Williams a, ) , Bob Boldi b , Anne Matlin b , Mark Weber b , Steve Hodanish c , Dave Sharp c , Steve Goodman d , Ravi Raghavan d , Dennis Buechler d a Parsons Laboratory, Massachusetts Institute of Technology, 77 Massachusetts AÕe., Cambridge, MA 02139, USA b Lincoln Laboratory, Massachusetts Institute of Technology, Lexington, MA 02420-9185, USA c National Weather SerÕice, Melbourne, FL 32935, USA d NASA Marshall Space Flight Center, HuntsÕille, AL 35806, USA Abstract Ž The development of a new observational system called LISDAD Lightning Imaging Sensor . Demonstration and Display has enabled a study of severe weather in central Florida. The total flash rates for storms verified to be severe are found to exceed 60 fpm, with some values reaching 500 fpm. Similar to earlier results for thunderstorm microbursts, the peak flash rate precedes the severe weather at the ground by 5–20 min. A distinguishing feature of severe storms is the presence of lightning ‘jumps’ — abrupt increases in flash rate in advance of the maximum rate for the storm. The systematic total lightning precursor to severe weather of all kinds — wind, hail, tornadoes — is interpreted in terms of the updraft that sows the seeds aloft for severe weather at the surface and simultaneously stimulates the ice microphysics that drives the intracloud lightning activity. q 1999 Elsevier Science B.V. All rights reserved. Keywords: Total lightning activity; Thunderstorms; Florida 1. Introduction This study is concerned with the electrification of severe weather, an appropriate topic for this special issue in honor of Bernard Vonnegut. The first examination of ) Corresponding author. Fax: q1-781-981-0632; e-mail: [email protected], [email protected]0169-8095r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. Ž . PII: S0169-8095 99 00011-3
Transcript
Ž .Atmospheric Research 51 1999 245–265
The behavior of total lightning activity in severeFlorida thunderstorms
Earle Williams a,), Bob Boldi b, Anne Matlin b, Mark Weber b,Steve Hodanish c, Dave Sharp c, Steve Goodman d,
Ravi Raghavan d, Dennis Buechler d
a Parsons Laboratory, Massachusetts Institute of Technology, 77 Massachusetts AÕe., Cambridge, MA 02139,USA
b Lincoln Laboratory, Massachusetts Institute of Technology, Lexington, MA 02420-9185, USAc National Weather SerÕice, Melbourne, FL 32935, USA
d NASA Marshall Space Flight Center, HuntsÕille, AL 35806, USA
Abstract
ŽThe development of a new observational system called LISDAD Lightning Imaging Sensor.Demonstration and Display has enabled a study of severe weather in central Florida. The total
flash rates for storms verified to be severe are found to exceed 60 fpm, with some values reaching500 fpm. Similar to earlier results for thunderstorm microbursts, the peak flash rate precedes thesevere weather at the ground by 5–20 min. A distinguishing feature of severe storms is thepresence of lightning ‘jumps’ — abrupt increases in flash rate in advance of the maximum rate forthe storm. The systematic total lightning precursor to severe weather of all kinds — wind, hail,tornadoes — is interpreted in terms of the updraft that sows the seeds aloft for severe weather atthe surface and simultaneously stimulates the ice microphysics that drives the intracloud lightningactivity. q 1999 Elsevier Science B.V. All rights reserved.
Keywords: Total lightning activity; Thunderstorms; Florida
1. Introduction
This study is concerned with the electrification of severe weather, an appropriatetopic for this special issue in honor of Bernard Vonnegut. The first examination of
0169-8095r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0169-8095 99 00011-3
( )E. Williams et al.rAtmospheric Research 51 1999 245–265246
electrification in a tornadic supercell storm is found in ‘‘Giant Electrical Storms’’Ž .Vonnegut and Moore, 1958 , a work inspired by Vonnegut’s personal observations ofthe renowned Worcester, MA storm in June 1953. This event strongly influencedVonnegut’s career as a scientist, as it stimulated his early thinking about the role of
Ž .convection in the electrification of storms Vonnegut, 1953 and the relationshipŽ . Ž .between electricity and tornadoes Vonnegut, 1960 . Vonnegut and Moore 1958 also
drew important attention to issues that remain with us today in the context of severeŽ .thunderstorms: 1 the extraordinarily high flash rates dominated by intracloud lightning;
Žin Vonnegut’s words, the Worcester storm was ‘‘going like gangbusters’’ as it went out. Ž . Ž .to sea late that evening ; 2 the extraordinary updraft velocities )100 mrs inferred
Ž .from simple parcel theory considerations; 3 the possible inconsistency between theŽ .observed radar cloud top height and conventional pseudoadiabatic parcel theory; 4 the
evidence for electrification and lightning in a large region of the upper storm, likelydevoid of supercooled water — an essential ingredient for the presently favored
Ž .precipitation mechanism for thunderstorm electrification; and 5 the possibility of anegatively charged cloud top in this superlative storm. Several of these issues will berevisited later in this paper.
Ž .The Worcester storm studied by Vonnegut and Moore 1958 was a major event in1953 that together focused national attention on severe weather and its formal definitionŽ .Galway, 1989 . Today, severe weather is defined by specific thresholds in wind, hailsize and vorticity. All of these phenomena have close physical connections with verticaldrafts in deep convection, that are themselves not directly measured with scanningDoppler weather radars. Cloud electrification and lightning are particularly sensitive tothese drafts because they modulate the supply of supercooled water that is the growth
Ž .agent for the ice particles ice crystals, graupel and hail believed essential for electricalcharge separation. For these reasons, one can expect correlations at the outset betweenlightning activity and the development of severe weather that may aid in understandingand predicting these extreme weather conditions. The exploration of these ideas histori-cally has been impeded by lack of good quantitative observations. A recent review of
Ž .results on severe storm electrification Williams, 1998a,b indicates a general absence ofcases for which total lightning activity is documented over the lifetime of a severe
Žstorm. The recent development of LISDAD Lightning Imaging Sensor Data Application. Ž .Display Boldi et al., 1998; Weber et al., 1998 has largely remedied this problem. The
LISDAD has been used in central Florida to quantify the behavior of total lightning inall types of severe weather.
2. Formal severe weather criteria and their connection with vertical drafts
Severe weather is characterized by at least one of the following three conditions,Ž .according to present National Weather Service criteria: 1 hailstones on the ground with
Ž .effective diameters greater than 0.75 in.; 2 a sustained surface wind in excess of 50Ž .knots; and 3 the occurrence of a tornado. All of these surface conditions have their
seeds in vertical storm drafts, the quantity most elusive to direct observations by
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 247
Doppler radar but a quantity strongly connected with cloud electrification and lightning.The systematic behavior of total lightning aloft relative to severe weather at the groundin this study warrants some discussion of these physical relationships.
2.1. SeÕere hail
Hail growth relies on particle levitation in a vertical airstream of supercooled water.Some estimates of the updraft strength required for hailstones of various diameters istherefore provided by the computation of the hailstone fall speed. Results in Fig. 1indicate that a vertical velocity of 29 mrs is needed to levitate a hailstone with the
Žcritical 3r4-in diameter. Fortuitously, this air speed is very close to the severe wind.speed of 50 knots to be addressed in the Section 2.2. The reduction in size due to
melting in the fall to the ground from the 08C isotherm will obviously require still largerdrafts aloft to account for the critical size at the ground.
2.2. SeÕere wind
Extreme wind at the surface in the vicinity of thunderstorms is often the result of adowndraft aloft. Mechanisms for downdrafts — gravitational loading by precipitationand cooling by evaporation and melting of condensate — have their origins in theupdraft and are expected to be enhanced by stronger updrafts. The observed tendency for
Žintracloud lightning to precede thunderstorm microbursts Goodman et al., 1988;.Williams et al., 1989a,b; Malherbe et al., 1992; Stanley et al., 1997 is consistent with
this general scenario. It is important to note, however, that the great majority ofmicroburst winds do not exceed the formal 50-knot criterion and, hence, are not
Ž .formally severe Williams, 1998a,b .
Fig. 1. Fall speed of ice spheres vs. sphere diameter at an altitude of 6 km MSL.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265248
Fig. 2. Illustration of the role of vertical drafts in vortex stretching.
2.3. Tornadoes
Tornadoes are intense vortices with a dominant vertical component of angularmomentum. Despite numerous theories for tornadogenesis, one feature common to all isthe vertical stretching of vorticity that is modulated by the vertical gradient of vertical
Ž .draft speed w i.e., dwrd z , as illustrated schematically in Fig. 2. For severe stormswhose vertical scale is strongly constrained by the tropopause, vertical stretching will belargely controlled by the magnitude of the drafts. Evidence will be presented later in thispaper that both updrafts and downdrafts are stretching vertical vorticity.
3. Methodology
The observational mainstay of this study is the LISDAD system in central Florida.The original intent of LISDAD was a ground-truthing system for optically-detectedlightning flashes from space using NASA’s Optical Transient Detector and the Light-ning Imaging Sensor. The flurry of severe weather in Florida in the spring and summerof 1997 soon made clear LISDAD’s effectiveness as a tool to study severe thunder-
Ž .storms Raghavan et al., 1997; Weber et al., 1998 . This currently operational real-timesystem integrates information from the prototype Integrated Terminal Weather SystemŽ .ITWS , developed by Lincoln Laboratory for the Federal Aviation Administration and
Ž .located in Orlando; the National Weather Service NWS WSR 88-D radar at Mel-Ž .bourne; the Storm Cell Identification and Tracking SCIT algorithm developed by the
Ž .National Severe Storms Laboratory Johnson et al., 1998 ; the Lightning Detection andŽ . Ž .Ranging LDAR system at the Kennedy Space Center Lennon and Maier, 1991 ; and
Ž . Ž .the National Lightning Detection Network NLDN Cummins et al., 1998 . TheLISDAD system offers substantial improvements over the traditional short-term fieldexperiment in the investigation of thunderstorms. The real-time, round-the-clock opera-
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 249
tion virtually guarantees capture of all interesting events. Furthermore, the directexposure and use by operational NWS forecasters provides insights about systematicfeatures of the observations as they occur. Finally, the different data sets that were ratherlaboriously assembled in the traditional field experiment after the fact are now availablefor integrated replay and inspection immediately following the events of interest.
The emphasis on total lightning as a diagnostic for severe weather in the LISDADresults gives the LDAR radiation data special importance. The viability of LDAR foraccurately detecting and mapping both intracloud and cloud-to-ground lightning flasheshas been verified through more than 25 years of operation at the NASA Kennedy Space
Ž . ŽCenter KSC . Its successful use during the TRIP Thunderstorm Research International. Ž .Program in the 1970s Lhermitte and Krehbiel, 1979; Lhermitte and Williams, 1985
demonstrated 50–100 m RMS errors in source locations for storms directly over KSC,based on observations from two independent arrays of radio receivers. More recentstudies in Orlando with the 3D lightning interferometer operated by the Office National
Ž . Ž .d’Etudes and de Recherches Aerospatiale ONERA Mazur et al., 1997 demonstratereliable detection of lightning at a range of 50 km, though with an attendant degradationof location accuracy. Some LDAR radiation is detected from storms on Florida’s westcoast at distances from KSC exceeding 200 km. For the rapidly migrating mesocyclonesof interest in this study, analysis to distances up to about 100 km from KSC will beconsidered.
The LDAR data stream currently ingested by LISDAD consists of individual radioŽ .source locations x, y, z,t that have been independently verified by the two independent
arrays of receivers at KSC. This data stream is used to create an LDAR flash rate, ameasure of the total flash rate for individual thunderstorm cells identified by SCIT. Inthis procedure, any source that occurs within 300 ms and 5000 m of a previous source isplaced into the same flash as the previous source. A flash can remain active for up to
Ž .5 s. Many of the flashes more than 10% are composed of just a single source. Suchflashes have been given the name ‘singletons’. The percentage of all LDAR flashes thatare singletons increases from 12 to 30% as the distance from the LDAR network to theflash increases from within 25 km to greater than 50 km.
The assignment of flashes to specific storm cells is identical for NLDN groundŽ .flashes and LDAR flashes: 1 advect the positions of the cells detected by the SCIT
algorithm to the current time using the ITWS track vectors provided for the respectiveŽ . Ž .cells; 2 assign the flash to all cells within 5 km of the flash location; and 3 if no cell
is found within 5 km, then assign the flash to the closest cell if that cell is within 35 kmof the flash location. Using these rules, about 95% of the flashes are assigned to a singlecell, with the remainder of the flashes being evenly split between zero and two cellassignments per flash. In examining the fast-moving supercells discussed in this paper, it
Ž . Ž .was discovered that rule 1 cell advection has a large influence on the computedminute-to-minute flash rates when cells move a distance about equal to their mean
Ž . Ž .intercell spacing 10 km in the time required for the NEXRAD radar update 5 min .For more detailed analysis of the storm structure in the vertical beyond the real-time
processing capability of LISDAD, the original Melbourne Doppler radar data have beenanalyzed after the fact. This includes the hand extraction of maximum reflectivity andmesocyclonic velocity on a tilt-by-tilt basis.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265250
All truth on severe weather otherwise documented with LISDAD remote sensing isbased on observer reports. This aspect of the study is judged to be the least quantitativeand most susceptible to sampling limitations. Information available in the MelbourneNWS office suggests that only about 10% of all storm data entries are accurate to within1 min of actual occurrence. Another 15% of all entries are accurate to within 2 to 5 min.
Ž .The largest percentage of entries 50% are accurate to within 6 to 10 min. The moresignificant the event, the more accurate are the reports.
4. General results
Although the focus of this study is on all types of severe weather in central Florida, itis useful to begin with some more general results from LISDAD that pertain to ordinaryŽ .non-severe thunderstorms as well as the broad spectrum of severe weather in allseasons. The use of the same rules to compute total flash rates in all thunderstormsregardless of their size and severity helps to place the results for extreme instability andshear in context.
Ž .The pop-up box feature in LISDAD Boldi et al., 1998 has been used to study thelightning histories of numerous Florida thunderstorms of all types. Severe thunderstorms
Žhave been identified on the basis of surface observer reports of hail dime size or. Ž .greater , strong wind trees blown down , or the occurrence of a tornado. Fig. 3
Ž .summarizes the peak flash rates LDAR for total lightning for all cases. The most likelymaximum flash rate, associated with small, non-severe thunderstorms in great abun-dance, is in the range of 1–10 per minute. A vertical dashed line is indicated at a flash
Ž .rate of 60 fpm 1 flash per second . To a large extent, the storms organize themselves
Fig. 3. Peak flash rates for Florida thunderstorms based on LDAR observations.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 251
Žinto non-severe and severe categories on the basis of peak flash rate alone with one.important caveat to be discussed presently . No severe cases were found with a peak
flash rate less than 60 fpm. For higher flash rates, the majority of cases were identifiedŽ .as severe. However, numerous cases with high flash rates one as high as 500 fpm were
found with no confirmation of severe conditions. Some of these high flash rate casesŽ .occurred over sparsely populated areas where hail for example may have been missed.
A few cases of high-flash-rate storms over heavily populated areas suggest that severestorm status was not attained.
The largest LDAR flash rates observed are in the vicinity of 500–600 fpm. The twodry season supercells discussed in Section 5 both lie in this tail of the flash ratedistribution in Fig. 3. Because both these storms were quite distant from the LDARnetwork for much of their lifetime, the uncertainty in the computed absolute flash ratesis larger than usual.
The fraction of thunderstorms found to be severe in Fig. 3 is surely larger than onemight find climatologically in Florida. This disproportionality is the result of theemphasis given to severe weather cases when a systematic behavior in the flash rateevolution became apparent in the early LISDAD observations.
The most obvious and systematic characteristic of severe thunderstorms is the rapidincrease in intracloud flash rate 1–15 min in advance of the severe weather manifesta-tion at the ground. These increases, termed lightning ‘jumps’, vary in magnitude from
Žabout 20 to over 100 fpmrmin. Examples of jumps in specific severe thunderstorms.can be seen in Figs. 6–8. The precursory nature of the lightning jump appears to pertain
not just to hail but to all severe weather, including strong wind and tornadoes. Aschematic history of total flash rate for a severe Florida thunderstorm is shown in Fig. 4
Ž .where three characteristic times t , t and t are shown. Time t marks the lightning0 1 2 0
jump, t the peak LDAR flash rate, and t the severe weather on the ground. A1 2
summary of such values for a wide range of Florida severe storm cases is shown inTable 1. On average, the recorded values are consistent with the evolution depicted in
Ž .Fig. 4. This table also includes values for peak flash rate LDAR and NLDN and
Fig. 4. Schematic lightning history in a Florida thunderstorm with t s jump time; t speak flash rate; and0 1
t s time of severe weather. Based on calculations of mean time differences from Table 1, if t s0, then2 0
t s7.4 min and t s16 min.1 2
( )E. Williams et al.rAtmospheric Research 51 1999 245–265252
Table 1Ž .LISDAD severe storm summary, t s time of rapid increase in LDAR flash rate the lightning ‘jump’ ,0
t s time of peak LDAR flash rate, t s time of first observer report of storm severity1 2
Ž . Ž . Ž .Severe NSSL SCIT Total LDAR Cloud-to- t UT t UT t UT0 1 2
weather cell IDa lightning: lightning grounddescription peak ‘jump’ lightning:
estimates of the magnitudes of the precursory lightning jumps. The majority of theŽ .information in Table 1 exclusive of surface observer reports was obtained by playback
Ž .of individual cases initially identified by NWS or Lincoln Laboratory ITWS personnel.The random error associated with surface observer reports of t was discussed in2
Section 3. The systematic error is not known. It seems likely, however, that someŽ .contribution to the mean lead times t y t , t y t are the lags in the reporting of the0 2 1 2
events.The existence of lightning jumps in the LDAR flash rate evolution is perhaps the
most obvious departure from steady-state behavior for the severe thunderstorms studied.The noted association between enhanced electrification and the growth of ice particlesaloft in the mixed-phase environment would suggest that the jumps are an accompani-ment of strong upsurges in air motion aloft. LISDAD evidence supports the idea that the
Fig. 5. Lightning ‘jump’ vs. maximum hailstone size. These results are drawn from Table 1.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265254
upsurges are linked with the growth of large hail. Fig. 5 shows the magnitude of thelightning jump vs. the maximum hailstone diameter reported on the ground for all hailcases in Table 1 observed with LISDAD. The positive correlation here supports aphysical connection with stronger electrification associated with stronger upsurges andlarger hail. Rough extrapolation downward to the millimeter-sized graupel characteristicof ordinary non-severe thunderclouds suggests lightning jumps less than 10 fpmrmin,consistent with observations.
The systematic flurry of intracloud lightning activity prior to tornadoes and water-Ž . Ž .spouts in this study Table 1 and in Goodman et al., 1998 is not without precedent.
Taylor identified peak intracloud flash rates 10–15 min prior to some tornadoes in theŽ . Ž .1970s W. Taylor, personal communication, 1996 . MacGorman 1993 documented a
maximum in intracloud lightning prior to the Binger tornado in 1986. Buechler et al.Ž .1996 noted a pronounced flurry of intracloud activity prior to a tornado touchdown inthe optical observations from HASA’s space-borne Optical Transient Detector. Thesimilar lightning signatures for both tornadoes and several waterspouts in this study leadus to draw no particular physical distinction between these two phenomena.
5. Case studies
The systematic evolution of events depicted schematically in Fig. 4 is now demon-Ž .strated for three specific cases drawn from Table 1: a hail case May 22, 1997 , a wind
Ž . Ž .case October 31, 1997 and a tornado case February 23, 1998 . The purpose of thesecomparisons is further clarification of the physical basis of the precursor signals in totallightning.
The evolution of total flash rate and maximum differential velocity at low levels forthe May 22, 1997 Orlando hail storm are shown in Fig. 6. Isolated convection developedshortly after noon local time to the northwest of Orlando International Airport on thisday. Within the next hour, new growth took place throughout the terminal area. Thestorm in question was too far from Melbourne to disclose the outflow history with the
Ž .NEXRAD radar, and so the Orlando Terminal Doppler Weather Radar TDWR wasused for this purpose. Richard Ferris, the ITWS site manager, observed oblate hailstoneswith diameters in the range of 3r4–1 in. at the site in the interval 1847–1852 UT, as
Ž .shown in Fig. 6. The strongest outflow of the day 72 knots was recorded by theTDWR at 1856 UT within 8 km of Ferris’s location. This storm therefore took on severestatus on the basis of both the hail and the microburst wind.
The lightning ‘jump’ phenomenon was recorded by LISDAD prior to both mi-Ž . Ž .crobursts at 1821 and 1838 UT , with the second, larger jump 75 fpmrmin preceding
the arrival of hail by about 9 min. It is interesting that the large hail precedes themaximum outflow by 4 to 7 min, a possible suggestion that the loading and meltingeffects of the smaller-size precipitation are playing the major role in forcing themicroburst, and the large hail fell out early on account of its significantly larger fallspeed. The 7-min lead times between peak flash rate and peak outflow agree very well
Žwith results for non-severe storms Goodman et al., 1988; Williams et al., 1989a,b;.Laroche et al., 1991; Malherbe et al., 1992; Stanley et al., 1997 , suggesting a similar
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 255
Ž .Fig. 6. May 22, 1997 hailstorm with severe microburst near Orlando. a History of total lightning flash rate,Ž .and b history of differential radial Doppler velocity at the surface.
physical basis for the precursor in both types of storms. The peak LDAR flash rate priorto the hail and large microburst is 275 fpm, substantially larger than values characteristic
Ž .for non-severe storms Fig. 3 .
( )E. Williams et al.rAtmospheric Research 51 1999 245–265256
Table 2Comparison of selected parameters for two florida supercells
Parameter 31 October 97 23 February 98
Pseudoadiabatic CAPE 1540 jrkg 2140 jrkgTropopause height 12.9 km 12.7 kmMelting level height 4.0 km 3.8 kmMaximum LDAR flash rate 554 fpm 567 fpmMaximum NLDN flash rate 14 fpm 17 fpmLightning ‘jump’ 60 fpmrmin 160 fpmrminMaximum ICrCG ratio ;230 ;200Maximum radar cloud top 16–17 km 17–18 kmTropopause overshoot 3–4 km 4–5 kmInferred maximum updraft speed 60–80 mrs 80–100 mrsDiameter of 30 dBZ core at 7 km 22 km 30 km
2 2 2 2Ž .Helicity 0–3 km 184 m rs 350 m rsMesocyclone maximum rotational velocity 19 mrs 28 mrsTypical mid-level mesocyclone diameter 5–9 km 5–8 kmSupercell translational speed 50–80 kmrh 90–100 kmrh
Ž .Hail ? no hail reported 3r4-in. hailŽ . Ž . Ž .Tornado ? no wind damage only yes F3
The selection of case studies from Table 1 for wind and tornado manifestations ofŽ .severe weather has a twofold purpose in this study: 1 to explore the vertical
development of the storm and its connection with total lightning precursors to severeŽ .weather and 2 to shed further light on the distinction between supercells that do and do
not produce tornadoes, a long-standing problem both scientifically and operationallyŽ . Ž .Burgess et al., 1993 . Improved Doppler radar observations Burgess et al., 1993 haveled to the realization that the majority of supercell mesocyclones do not evolve totornadoes. A challenging issue is the identification of physical conditions that make thedifference. With this challenge in mind, two electrically extreme supercell mesocyclonesin the Florida dry season were selected from the LISDAD archive from Table 1 to
Ž .compare — one on February 23, 1998 that produced an F3 tornado and another onŽ .October 31, 1997 for which wind damage was reported, but no tornado . Selected
parameters for comparison of these two cases are shown in Table 2. Included in thisTable are values for tropopause overshoot and inferred maximum updraft speed,
Ž .following like calculations made initially by Vonnegut and Moore 1958 . The numbersare for the most part quite similar, thereby emphasizing the subtlety of the distinctionbetween supercells that do and do not produce tornadoes. For example, the peak LDARflash rates agree to within 10% and are both extraordinarily high. It is possible that the
Ž .use of the same non-severe storm rules leads to an overcounting of flashes. It is worthnoting, however, that both estimates are less than the value for stroke rate estimated by
Ž . Ž .Fig. 7. October 31, 1997 supercell with severe wind Polk County : a time–height plot of maximum radarŽ . Ž . Ž .reflectivity dBZ , b history of total lightning flash rate, c time–height plot of maximum mesocyclonic
Ž . Ž . Ž .rotational velocity mrs , d history of cloud-to-ground flash rate, and e Florida map showing supercellstorm track.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 257
( )E. Williams et al.rAtmospheric Research 51 1999 245–265258
Ž . ŽVonnegut and Moore 1958 for the Worcester, MA tornadic storm 600–1200.strokesrmin , the only visual observation of stroke rate in a nighttime tornadic
supercell.The two parameters in Table 2 showing the largest contrast between the tornadic and
non-tornadic supercell are the lightning jump and the helicity in the 0–3 km heightrange. In both categories, the larger value is associated with the tornado-producer onFebruary 23.
Histories of radar reflectivity and mesocyclonic rotational velocity in time–heightŽ .format, together with the lightning LDAR and NLDN ground flashes evolutions for the
two cases are shown in Figs. 7 and 8. The storm intervals containing the largestlightning jump, maximum flash rate and most intense vertical development are includedin both cases, and the overall storm tracks are also shown in Figs. 7 and 8. Themagnitude of the lightning jump showed the largest contrast between the two casesamong all parameters in Table 2, with the tornado-producing case showing a substan-
Ž .tially larger value 160 fpmrmin . Neither storm was sufficiently close to the MelbourneŽ .radar to enable observation of concentrated low-level vorticity i.e., the tornado . These
time–height comparisons reveal substantially more about the differences between thetwo cases than the parameter comparisons in Table 2.
ŽIn comparison with ordinary airmass thunderstorms Lhermitte and Krehbiel, 1979;.Lhermitte and Williams, 1985; Goodman et al., 1988; Williams et al., 1989a,b , these
supercells are far closer to a dynamical steady state in their vertical development. Andyet, the observations show in both cases that the unsteady features are of centralimportance in signaling severe conditions on the ground. The total lightning is perhaps
Žthe least steady feature of supercell evolution, with substantial lightning jumps in the. ŽLDAR flash rate coinciding with explosive vertical development 2020–2040 UT on
.October 31 and 0305–0320 UT on February 23 that are again precursory to severeŽ .weather at 2045 UT on October 31 and at 0355 UT on February 23 . The upward
Žgrowth at mid-levels and in particular in the mixed-phase region where the strongest.charge separation is expected clearly coincides with the enhancement in mid-level
rotation, presumably by stretching of vertical vorticity in the updraft at mid- and upperlevels. In both cases, the maximum in rotational velocity aloft is sustained to the time of
Ž .maximum LDAR flash rate 2045 UT on October 31 and 0324 UT on February 23 .ŽUnlike the behavior of many non-severe thunderstorms Byers and Braham, 1949;
.Williams, 1985a , the peak flash rate does not coincide with the maximum radar cloudtop height. Agreement is better between the vertical extent of radar reflectivity in themixed-phase region at lower levels, consistent with the idea that supercooled water is afundamental ingredient in the electrification process. This leaves unresolved the question
Ž .raised by Vonnegut and Moore 1958 concerning the electrical role of the large quantityof ice particles at altitudes above the mixed-phase region in supercell storms. Present
Ž . Ž .Fig. 8. February 23, 1997 supercell with F3 tornado Volusia County : a time–height plot of maximum radarŽ . Ž . Ž .reflectivity dBZ , b history of total lightning flash rate, c time–height plot of maximum mesocyclonic
Ž . Ž . Ž .rotational velocity mrs , d history of cloud-to-ground flash rate, and e Florida map showing supercellstorm track.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 259
( )E. Williams et al.rAtmospheric Research 51 1999 245–265260
Ž .laboratory simulations Takahaski, 1978; Saunders et al., 1991 do not extend to thistemperature region. Following the maximum in total flash rate in both cases, an abruptdrop in flash rate occurs, suggesting a reduction in updraft strength with an attendantreduction in rotational velocity. Severe wind is reported at the surface for the October 31case, but no severe report was logged for February 23 at the equivalent time.
Shortly after the abrupt diminishments in flash rate, in both cases, a secondaryŽmaximum in rotational velocity is observed 2057 UT on October 31 and 0337 UT on
. ŽFebruary 23 associated with the most strongly descending reflectivity contours and.declining reflectivity within the respective mesocyclonic cores , indicative of possible
restretching of the vorticity, but in this case by a downdraft rather than an updraft.At this juncture, the behavior of the two cases diverges. The flash rate for October 31
rebuilds after its short-term decline, whereas the flash rate for February 23, that hasdropped to a lower relative level, does not recover. The mid-level reflectivity forOctober 31 is sustained, whereas for February 23 at mid level, reflectivity contours
Ž .continue the descent that began near the time of cloud top apogee 0310 UT . An F3Žtornado is observed at 0355 UT with a notable diminishment of rotational velocity
.aloft in the latter case, but no further severe weather is observed in the time frameshown for the October 31 supercell.
The lightning discussion has thus far centered on the LDAR information on accountof the demonstrated connection with storm vertical development. The sustained lightningjumps that stand out clearly in the LDAR history are hardly present in the NLDNground flash history, and the general level of activity is less than the inferred intraclouddevelopment, often by more than 10-fold. Some tendency is noted for suppressed ground
Žflash activity at times of elevated intracloud activity 2040 UT on October 31 and 0330.UT on February 23 , suggesting a competition between lightning types for a common
The flash rates and lightning ‘jumps’ recorded in Fig. 3 and in Table 2 in severeFlorida thunderstorms are extraordinarily large in comparison with ordinary non-severestorms. This possibly controversial result and the general strategy in defining a storm’slightning activity therefore deserve some discussion. A central issue in this context is thephysical nature of a lightning flash. Though complicated and still poorly understoodŽBernard Vonnegut’s pointed quote on this topic: ‘‘What theorist would have predicted
.lightning?’’ , the lightning flash is a well-defined physical entity, supported by numer-ous measurements. A flash is a connected plasma whose electrical conductivity iseverywhere larger than the air dielectric in which it is embedded. By ‘connected’ wemean that at any instant in the flash’s lifetime, every pair of points within the flash arelinked by some path with elevated electrical conductivity. For optical measurementswith limited sensitivity, a flash may appear to cease in the dark interstroke interval.
ŽHowever, evidence from field change measurements and radar Hewitt, 1957; Williams.et al., 1989a,b support the idea that electrical current continues to flow during this
Ž .interval and the flash is sustained. As noted by Heckman and Williams 1989 and
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 261
Ž .Mazur et al. 1997 , continued growth of the flash somewhere in space into thesurrounding electric field is required for flash sustenance.
Problems naturally arise when a flash is to be documented with inadequately sampledoptical or RF measurements, such as the LDAR maps of flashes discussed here. If thelightning flash radiated continuously above the measurement noise level as it progressedin space, and if the measurement sampling were continuous in time, the accuratedepiction of all flashes following the above definition would be straightforward. Inpractice, lightning does not radiate strongly at all times, and furthermore, the LDARsampling and processing is incomplete, with a maximum sample rate of 10 kHz. These
Žlimitations force the selection of rather simple space and time criteria but nonetheless,criteria consistent with statistical information on the durations and extents of lightning
.flashes in defining flashes as described in Section 3.This sampling limitation for flash definition is strongly aggravated when the flash
rate increases to the point where the interflash interval is comparable to or less than theflash duration and overlap in space and time is prevalent. This study has shown that thiscondition is very common in severe weather. This difficulty with the interpretation ofLDAR radiation sources has also appeared recently in the analysis of optical pulses
Ž . Žobserved by NASA’s Lightning Imaging Sensor LIS in space Christian and Boccip-.pio, 1998; personal communication and in Oklahoma supercells observed by a portable
Ž .LDAR system Krehbiel, 1998; personal communication . These problems are bestappreciated with some simple pictures of flashes within clouds.
Three possible scenarios for the occurrence of lightning flashes in active storm ‘cells’Žare illustrated in Fig. 9. Here, we have assumed for lack of a more strongly supported
. Ž .alternative that every lightning flash is a double-ended ‘tree’ Mazur et al., 1997 . Fig.Ž .9 a depicts a situation with a single flash without overlap in either space or time. Fig.Ž .9 b shows flash activity in different regions of space but overlapping in time. Finally,
Ž .Fig. 9 c shows a scenario with overlapping flashes in both space and time.Ž .Fig. 9 a is the usual picture for an ordinary non-severe thunderstorm in which the
charging zone and breakdown region are highly localized and in which the interflashinterval is larger than the flash duration. An electrically stressed spark gap presents thesame situation. Overlapping flashes are impossible in such a case, as breakdownprevents the occurrence of new breakdown to form the next flash. One possible flaw in
Ž . Ž .Fig. 9. Lightning configurations in a small storm with no flash overlap in space or time, b severe stormŽ .with flash overlap in time but not in space, and c severe storm with flash overlap in space and in time.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265262
the grouping algorithm described in Section 3 is the inadvertent decomposition of singlelarge flashes into multiple smaller ones.
Ž .Fig. 9 b shows a possible scenario for a high flash rate supercell with extremelyvariable conditions in both particle charging and in dielectric strength, that leads tosimultaneous flashes in different regions. In storms of this kind characterized byextremes in draft strength, liquid water content and ice particle size, the stronglyheterogeneous conditions in microphysical growth are obvious. If such conditions are
Ž .linked with particle charging Baker et al., 1987; Williams et al., 1991, 1994 , then wecan expect heterogeneous charging. If large ice particles are influential in weakening thedielectric strength of the upper storm and aiding in the initiation of lightning flashes, wecan expect multiple breakdown. The SCIT cell identification procedure in this study
Ž .frequently identifies the entire supercell as one ‘cell’ with one pop-up box history .Ž .These cells are often 20–30 km in diameter Table 2 and 15–17 km deep. The
Ž .substructure of these cells is readily apparent in the radar observations not shown .w Ž .xLightning flashes that overlap in space as in Fig. 9 c may not seem a likely
possibility for reasons advanced in the earlier discussion of the spark gap. However, thethunderstorm is a continuum dielectric that is continuously charged. In the case of solid
Ž .dielectric materials Williams et al., 1985b charged continuously with an externalelectron beam, individual discharges occasionally overlapped in space. It is not clear towhat extent this happens in severe weather, but if it does, it would present the mostdifficult problem of flash distinction.
One could escape the flash overlap problem entirely by abandoning the grouping ofLDAR sources into flashes described in Section 3 and monitoring the LDAR sourcesalone as a measure of storm electrical activity. We have not followed this approach forone major reason: the characterization of the lightning activity is then dependent on the
Ž .sensitivity of the detection system LDAR, in the present context and on the distancebetween lightning and detector. If the system is sufficiently sensitive to record only onesource per flash, then we achieve the desired total flash rate. A characterization oflightning activity that is independent of measurement system is clearly preferable. Most
w Ž .previous results on electrical activity in severe weather reviewed by MacGorman 1993Ž .xand Williams 1998a,b are difficult to compare quantitatively with the present results
because insufficient information is available on the physical process being monitoredŽand counted. This problem may not be so serious for an optical sensor in space e.g., the
.Lightning Imaging Sensor that maps optical pulses from lightning in storms all over theworld, but is almost surely a problem for surface measurements on different storms indifferent locales.
One objection to the use of ‘flash’ in quantifying lightning activity is the variablenature of this unit in energy and charge moment, to name just two physical properties.Indeed, the charge moments of flashes we define by LDAR groupings may vary from a
Ž .fraction of a coulomb-kilometer for delicate intracloud flashes high in the storm toseveral thousand coulomb-kilometers for the large ‘spider’ lightning in mesoscale
Ž .convective systems Williams, 1998a,b; Huang et al., 1998 . We view this variability asan inevitable consequence of the meteorological variability. Furthermore, the history oflightning studies gives us a much better feel for energy and charge moment of flashesthan for the physical processes with which individual LDAR sources may be associated.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 263
For all of these reasons, we prefer to quantify electrical activity as a flash rate in thesevere weather cases examined. A single rule for grouping LDAR sources into flashes isapplied in all cases. The flash rates we obtain for non-severe thunderstorms are
Ž .reasonable ones Fig. 3 . Though errors may arise due to spacertime overlap in thesevere regime when the flash rates are quite high, the values obtained are notimplausible.
7. Conclusions
The LISDAD system has revealed a remarkably consistent pattern of total lightningbehavior for severe Florida thunderstorms, with strong upsurges prior to severe weather
Ž .in all categories wind, hail and tornadoes , in both the wet and the dry seasons. Theupdraft appears to be causal to both the extraordinary intracloud lightning rates and thephysical origin aloft of the severe weather at the surface. The supercell comparison has
Ždisclosed deep reservoirs of vertical mesocyclonic angular momentum to 10 km.altitude , with indications of vortex stretching by both updrafts initially and by down-
drafts at later stages. These cases and additional tornadorwaterspout cases considered inŽ . Ž .greater detail by Goodman et al. 1998 and Hodanish et al. 1998 are consistent in
showing that pronounced departures in dynamical steady state are needed for tornadoge-nesis. In particular, a slumping of the cloud and attendant diminishment in total flashrate after the initial lightning jump appear necessary to concentrate vorticity near the
Žsurface. Continued examination of Florida null cases i.e., mesocyclones without.tornadoes with the LISDAD are needed for further clarification of mechanisms.
8. Further Reading
The following references are also of interest to the reader: Perez et al., 1997,Rasmussen and Straka, 1996, Storm Data, 1997.
Acknowledgements
This work was sponsored by the National Aeronautics and Space Administration.Leslie Mahn prepared the manuscript with patience and precision. Discussions on severeweather with D. Burgess, C. Doswell, D. MacGorman R. Markson, E. Rasmussen andW. Taylor are greatly appreciated. Richard Ferris provided valuable on-site observationsof many of the Florida storms. P. MacKeen, T. Smith and M. Eilts provided guidance onthe NSSL algorithm. The LISDAD project has been supported by Jim Dodge of theNASA Earth Science Enterprise under Contract H-23725D. We thank Ralph Marksonfor additional support to Lincoln Laboratory on a NASA STTR project.
References
Baker, B., Baker, M.B., Jayaratne, E.R., Latham, J., Saunders, C.P.R., 1987. The influence of diffusionalgrowth rates on the charge transfer accompanying rebounding collisions between ice crystals and softhailstones. Quart. J. R. Meteorol. Soc. 113, 1193–1215.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265264
Boldi, R., Williams, E., Matlin, A., Weber, M., Hodanish, S., Sharp, D., Goodman, R., S., Raghavan, 1998.Ž .The design and evaluation of the Lightning Imaging Sensor Data Applications Display LISDAD . 19th
Conference on Severe Local Storms, AMS.Buechler, D.E., Blakeslee, R.J., Christian, H.J., Creasy, R., Driscoll, K., Goodman, S.J., Mach, D.M., 1996.
Lightning activity in a tornadic storm observed by the optical transient detector. Preprints 18th Conferenceon Severe Local Storms, American Meteorological Society, San Francisco, CA, February.
Burgess, D.W., Donaldson, R.J. Jr., P.R. Desrochers, 1993. Tornado detection and warning by radar, Thetornado: its structure, dynamics, predictions and hazards. In: Church, C., Burgess, D., Doswell, C.,
Ž .Davies-Jones, R. Eds. , AGU Geophysical Monograph, Vol. 79.Byers, H.R., Braham, R.R., 1949. The Thunderstorm, US Government Printing Office.
Ž .Christian, H., Boccippio, D., 1998 personal communication .Cummins, K.L., Murphy, M.J., Bardo, E.A., Hiscox, W.L., Pyle, R.B., Pifer, A.E., 1998. A combined
TOArMDF technology upgrade of the U.S. National Lightning Detection Network. J. Geophys. Res. 103,9035–9044.
Galway, J.G., 1989. The evolution of severe thunderstorm criteria within the weather service. WeatherForecast. 4, 585–592.
Goodman, S.J., Buechler, D.E., Wright, P.D., Rust, W.D., 1988. Lightning and precipitation history of amicroburst-producing storm. Geophys. Res. Lett. 15, 1185–1188.
Goodman, S., Raghavan, R., Williams, E., Boldi, R., Matlin, A., Weber, M., 1998. Total lightning and radarstorm characteristics associated with severe storms in central Florida AMS, 19th Conference on SevereLocal Storms.
Heckman, S.J., Williams, E.R., 1989. Corona envelopes and lightning currents. J. Geophys. Res. 94,13287–13294.
Hewitt, F.J., 1957. Radar echoes from interstroke processes in lightning. Proc. Phys. Soc. London B70,961–979.
Hodanish, S., Sharp, D., Williams, E., Boldi, R., Matlin, A., Weber, M., Goodman, S., Raghavan, R., 1998.Observations of total lightning associated with severe convection during the wet season in central Florida.AMS, 19th Conference on Severe Local Storms.
Huang, E.W., Williams, E., Boldi, R., Heckman, S., Lyons, W., Taylor, M., Wong, C., Nelson, T., 1998.Criteria for elves and sprites based on Schumann resonance observations. J. Geophys. Res.
Johnson, J.T., MacKeen, P.L., Witt, A.E., Mitchell, E.D., Stumpf, G., Thomas, K., 1998. The storm cellidentification and tracking algorithm: an enhanced WSR-88D algorithm. Weather Forecast. 13.
Laroche, P., Malherbe, C., Bondiou, A., Weber, M., Engholm, C., Coel, V., 1991. Lightning activity inmicroburst producing storms. 25th International Conference on Radar Meteorology, Paris, France.
Lennon, C., Maier, L., 1991. Lightning mapping system, NASA CP-3106, Vol. 2. International Aerospace andGround Conference on Lightning and Static Electricity, 89-1 to 89-10.
Lhermitte, R., Krehbiel, P., 1979. Doppler radar and radio observations of thunderstorms. IEEE Trans. Geosci.Electron. GEr17, 162–171.
Lhermitte, R., Williams, E., 1985. Thunderstorm electrification: a case study. J. Geophys. Res. 90, 6071–6078.MacGorman, D.R., 1993. Lightning in tornadic storms: a review. In: Church, C., Burgess, D., Doswell, C.,
Ž .Davies-Jones, R. Eds. , The Tornado: Its Structure, Dynamics, Prediction and Hazards, GeophysicalMonograph, Vol. 79, American Geophysical Union.
Malherbe, C., Pigere, J., Blanchet, P., Deste, O., Bondiou, A., Laroche, P., 1992. Relation entre l’activiteelectrique d’orage et le developpement de microbursts-experience. MITrONERA, Orlando, FL, 1991, RF´ONERA aB0516154P4.
Mazur, V., Williams, E., Boldi, R., Maier, L., Proctor, D., 1997. Initial comparison of lightning mapping withoperational time-of-arrival and interferometric systems. J. Geophys. Res. 102, 11071–11085.
Raghavan, R., Goodman, S. Meyer, P., Boldi, R., Matlin, A., Weber, M., Williams, E., Sharp, D., Hodanish,S., Madura, J., Lennon, C., 1997. A real-time examination of the incremental value of lightning data indiagnosing convective storm characteristics. AMS, Preprints, Seventh International Conference on Avia-tion, Range, and Aerospace Meteorology, Long Beach, CA.
( )E. Williams et al.rAtmospheric Research 51 1999 245–265 265
Rasmussen, E.N., Straka, J.M., 1996. Mobile mesonet observations of tornadoes during VORTEX-95. AMSConference on Severe Local Storms, 1–5, San Francisco, CA.
Saunders, C.P.R., Keith, W.D., Mitzeva, R.P., 1991. The effect of liquid water content on thunderstormchanging. J. Geophys. Res. 96, 11007–11017.
Stanley, M., Krehbiel, P., Rison, W., 1997. Lightning as a Precursor of Outflow and Downbursts fromThunderstorms. 28th Conference on Radar Meteorology, Austin, TX, American Meteorological Society,pp. 151–152.
Ž .Storm Data, 1997. NOAA Ashville, NC , Vol. 39, No. 10.Takahaski, T., 1978. Riming electrification as a charge generation mechanism in thunderstorms. J. Atmos.
Sci., pp. 1536–1548.Vonnegut, B., 1953. Possible mechanism for the formation of thunderstorm electricity. Bull. Am. Meteorol.
Soc. 34, 378.Vonnegut, B., 1960. Electrical theory of tornadoes. J. Geophys. Res. 65, 203–221.
Weber, M.E., Williams, E.R., Wolfson, M.M., Goodman, S.J., 1998. An assessment of the operational utilityof a GOES lightning mapping sensor. Project Report NOAA-18, MIT Lincoln Laboratory, Lexington, MA.
Ž .Williams, E.R., 1985a. Large scale charge separation in thunderstorms. J. Geophys. Res. 90 D4 , 6013–6025.Williams, E.R., 1989. The tripole structure of thunderstorms. J. Geophys. Res., 13151–13167.Williams, E.R., 1998a. The positive charge reservoir for sprite-producing lightning. J. Atmos. Sol. Terr. Phys.
60, 689–692.Ž .Williams, E.R., 1998b. The electrification of severe storms. In: Doswell C.A. Eds. , AMS Monograph on
Ž .Severe Storms in review .Williams, E.R., Cooke, C.M., Wright, K.A., 1985b. Electrical discharge propagation in and around space
charge clouds. J. Geophys. Res. 90, 6059–6070.Williams, E.R., Weber, M., Orville, R., 1989a. The relationship between lightning type and convective state of
thunderclouds. J. Geophys. Res. 94, 13213–13220.Williams, E.R., Geotis, S.G., Bhattacharya, A.B., 1989b. A radar study of the plasma and geometry of
lightning. J. Atmos. Sci. 46, 1173–1185.Williams, E.R., Zhang, R., Rydock, J., 1991. Mixed-phase microphysics and cloud electrification. J. Atmos.
Sci. 48, 2195–2203.Williams, E.R., Zhang, R., Boccippio, D., 1994. Microphysical growth state of ice particles and large-scale
electrical structure of clouds. J. Geophys. Res. 99, 10787–10793.
